Methods for creating Tuvalu demographic data
Wes Hinsley
8 November 2017
Introduction
Tuvalu, formerly known as the Ellice Islands, is a Polynesian island nation located in the Pacific Ocean, comprising three reef islands, and six atolls. It covers 26 square kilometres, and in the 2012 census, it had a population of 10,640.
Because of its small size, UNWPP only provides estimates and predictions for the total population, not gender-divided, and no other fields. For VIMC purposes however, we need the full suite of fields like any other country as part of the set of GAVI countries.
Tonga is the nearest country to Tuvalu for which UNWPP provides comprehensive data. This report describes the simple method of using Tonga’s data to create a proportional surrogate set of data for Tuvalu.
Method
For all of the data fields in UNWPP that are expressed as rates (mortality rate, birth rate, etc), we assume we can use UNWPP’s data for Tonga as applying to Tuvalu equivalently.
This also applies for the under-5 and under-1 child mortalities. We will then use IGME’s neonatal mortality data (which is available for Tuvalu) to produce a hybrid that is consistent with the rest of the data, while being guided by the real data we have. See the documentation for the child mortality methods elsewhere.
Where the data fields are absolute numbers of people though; (number of deaths, number of people of certain age, etc), we want to use simple scaling, where the scaling factor is the population of Tuvalu, divded by the population of Tonga, as it was in the latest estimate - 2015.
While using a single scaling factor across the time series seems crude, when we could scale in a year-by-year way, it is necessary to provide internal consistency with the other UNWPP data for Tonga. If we used the dynamic TUV/TON fraction across all years, then we would have a very accurate-looking total population, but the scaled rates of change for Tonga that vary through time would not be consistent with the total population. Therefore, scaling of the whole population with a single factor is the necessary approach for consistency.
Note that the methods below are performed after the population data is extended to included year-by-year detail for over-80 years olds, prior to 1990 - documented elsewhere. The Tuvalu data is created from Tonga data that has already been extended.
Total Population
Comparisons for scaled absolute numbers
The following graphs show for each statistic type represented by absolute numbers, (as opposed to rates), where we have scaled from Tonga’s data to approximate Tuvalu’s. We only have data on the total population to compare to.
Fertility
Population by age
UNWPP provide population data in two forms. The first, form is “quinquennial” (QQ), meaning the number of people in a 5-year age-band, reported at 5-year intervals. This is the resolution at which UNWPP’s source data is represented. The second form is “interpolated”, where UNWPP provide population data in 1-year age-bands, at single yearly intervals.
